Investigation of α-Synuclein Fibril Structure by Site-directed Spin Labeling*

The misfolding and fibril formation of α-synuclein plays an important role in neurodegenerative diseases such as Parkinson disease. Here we used electron paramagnetic resonance spectroscopy, together with site-directed spin labeling, to investigate the structural features of α-synuclein fibrils. We generated fibrils from a total of 83 different spin-labeled derivatives and observed single-line, exchange-narrowed EPR spectra for the majority of all sites located within the core region of α-synuclein fibrils. Such exchange narrowing requires the orbital overlap between multiple spin labels in close contact. The core region of α-synuclein fibrils must therefore be arranged in a parallel, in-register structure wherein same residues from different molecules are stacked on top of each other. This parallel, in-register core region extends from residue 36 to residue 98 and is tightly packed. Only a few sites within the core region, such as residues 62–67 located at the beginning of the NAC region, as well as the N- and C-terminal regions outside the core region, are significantly less ordered. Together with the accessibility measurements that suggest the location of potential β-sheet regions within the fibril, the data provide significant structural constraints for generating three-dimensional models. Furthermore, the data support the emerging view that parallel, in-register structure is a common feature shared by a number of naturally occurring amyloid fibrils.

The toxicity of ␣-synuclein and that of other amyloidogenic proteins has been linked to a misfolding process that involves a number of oligomeric species and ultimately results in fibril formation (22). While the precise mechanisms of toxicity remain to be fully elucidated, it is clear that some of the misfolded oligomeric species are highly cytotoxic. To better understand and ultimately prevent the misfolding process, it is important to decipher the underlying molecular mechanisms. Inasmuch as ␣-synuclein fibril formation represents the end point of the misfolding process and is considered a hallmark of the aforementioned diseases, the analysis of its fibril structure has received significant attention. Fiber diffraction studies have shown that fibril formation results in formation of the classical "cross-␤" structure (23). In this cross-␤ structure, individual ␤-strands are arranged perpendicular to the fibril axis spaced at distances of 4.7-4.8 Å apart from each other. A number of studies using electron paramagnetic resonance (EPR) spectroscopy, proteolysis, hydrogen-deuterium (H/D) exchange and solid-state nuclear magnetic resonance (ssNMR) spectroscopy (24 -27) have suggested that the cross-␤ core region is located in the central portion of ␣-synuclein. While this core region is highly structured, the N-and C-terminal regions appear to be less ordered, with the C terminus being largely unfolded. Surprisingly, it was found that the borders of the core region extend significantly beyond the NAC region (residues . Various studies suggest that the core region begins somewhere in the range of residues 31-39 and ends around residues 95-109 (24 -27). Despite these results and the initial assignment of secondary structure for selected parts of the core region by ssNMR (26), the precise extent of the core region and its three-dimensional structure remain unknown.
In the present study, we applied EPR spectroscopy, together with site-directed spin labeling (SDSL) (28 -30), to further investigate the structure of ␣-synuclein fibrils. We generated 83 different spin-labeled derivatives of ␣-synuclein, including a nitroxide scan wherein each residue of ␣-synuclein (from 30 -101) was replaced by spin label R1 (Fig. 1), one amino acid at a time. EPR analysis of fibrils grown from these derivatives was performed in order to: (a) define the precise location of the core region, (b) assign secondary structural elements within the core region, and (c) investigate the reported parallelism within ␣-synuclein fibrils (24).

EXPERIMENTAL PROCEDURES
Generation of ␣-Synuclein Single Cysteine Mutants-A truncation mutant (115ter) of human ␣-synuclein, containing residues 1-115, was generated by adding two stop codons between residues 115 and 116. The original wild-type human ␣-synuclein construct (in pRK172) that was used for this modification was provided by Dr. M. Goedert (Medical Research Council Laboratory, Cambridge, UK). No native cysteines exist in human ␣-synuclein. Single cysteines were introduced by sitedirected mutagenesis and verified by DNA sequencing.
Protein Expression and Purification of 115ter ␣-Synuclein-Proteins were expressed and purified based upon the previously described protocol for the full-length protein (24). The only notable difference was that the steps after acid precipitation and centrifugation were modified. At this point, the 115ter proteins were dialyzed against 20 mM MES, pH 6.0, 1 mM dithiothreitol, and 1 mM EDTA, then loaded onto a HiTrap SPXL column (Amersham Biosciences, GE Healthcare) that was equilibrated with the same buffer. Proteins were then eluted in 0 -1 M NaCl gradient, and fractions were analyzed by SDS-PAGE. Fractions containing 115ter ␣-synuclein were subsequently pooled and 1 mM dithiothreitol was added. Protein purity was Ͼ95% according to Bio-Safe TM Coomassie Stain (Bio-Rad). Protein concentration was determined by using a Micro BCA protein assay kit (Pierce).
Spin Labeling and Fibril Assembly of 115ter ␣-Synuclein-Recombinant ␣-synuclein was filtered through an Amicon YM-100 spin filter (MWCO 1 ϫ 10 5 , Millipore) to remove any pre-aggregates. Immediately before spin labeling, dithiothreitol was removed by loading the protein solution onto a PD-10 column (Amersham Biosciences, GE Healthcare) equilibrated with buffer containing 20 mM Hepes, pH 7.4, 100 mM NaCl, and 1 mM EDTA, then eluted using the same buffer. ␣-Synuclein was labeled with 10-fold molar excess of R1 MTSL spin label [1-oxyl-2,2,5,5-tetramethyl-D-pyrroline-3-methyl]methanethiosulfonate (Toronto Research Chemicals, Toronto, Ontario, Canada) or with a mixture of diluted R1 spin label and the diamagnetic analog, R1Ј [1-acetyl-2,2,5,5-tetramethyl-3-pyrroline-3methyl]-methanethiosulfonate (Toronto Research Chemicals) (see Fig. 1) for 1 h at room temperature. Excess label was removed using PD-10 columns with the aforementioned elution buffer. Spin-labeled proteins were washed twice with elution buffer, concentrated using Amicon Ultra-4 centrifugal filter units (MWCO 5 ϫ 10 3 , Millipore), and assembled into fibrils as described previously (24). Spin-labeled proteins (100 M or higher) were incubated in the aforementioned elution buffer at 37°C under constant agitation for 3-5 days. Fibrils were harvested by ultracentrifugation and washed twice with the same elution buffer. The quality of the fibrils was monitored by electron microscopy (EM) of negative-stained samples, by circular dichroism (CD) spectroscopy, and by thioflavin T fluorescence spectroscopy as described (31). The truncated synuclein formed fibrils much faster than did the full-length wild-type (8 h versus 48 h). Fibril dimensions were similar to that of wild-type, although increased lateral aggregation of the 115ter-based fibrils was noted. Seeding R1-labeled 115ter with full-length wild-type ␣-synuclein fibrils was performed as reported previously (32).
X-band EPR Spectroscopy and Data Analysis-The derivatized proteins were loaded into quartz capillaries (0.6-mm inner diameter ϫ 0.84-mm outer diameter, VitroCom, Mt. Lakes, NJ) or TPX capillaries (for accessibility measurements), and EPR spectra were recorded on X-band Bruker EMX spectrometers (Bruker Instruments, Billerica, MA). Full spectral scans were performed using an ER 4119HS resonator, and accessibility measurements were performed using an ER 4123D dielectric resonator at room temperature. The spectral scans obtained at room temperature using the ER 4119HS resonator had a scan width of 150 Gauss at an incident microwave power of 12 milliwatt. All EPR spectra shown were normalized to the same amount of spins, using double integration, and are presented as normalized. For distance measurements, EPR spectra were obtained at 233 K using a Bruker N 2 temperature controller (ER4131VT), and were recorded at 200 Gauss scan width. The O 2 accessibility ⌸ (O 2 ) of fibrils from 25% R1-labeled 115ter ␣-synuclein was determined using the standard power saturation method (33). EPR spectra were recorded as a function of microwave power (from 100.58 to 1.59 milliwatt) in the presence of pure O 2 and N 2 at room temperature. The peak-to-peak amplitude of the central line was plotted as a function of incident microwave power and fitted to obtain the P 1/2 parameters for O 2 and N 2 (33). The dimensionless quantity ⌸ (O 2 ) was obtained from the difference of the respective P 1/2 values divided by the peak-to-peak line width of the given sample. To correct for variations in resonator properties, we also applied the normalization using the diphenyldipicrylhydrazide standard as described in reference (33). Sample stability was verified in control experiments, which demonstrated that EPR spectra and accessibilities did not change over the time course of one month. Simulation of dipolar broadening was performed using software generously provided by Drs. Altenbach and Hubbell, as described previously (34).

Exchange Narrowing Reveals Parallel, In-register Structure with Physical Contact between the Same Residues in Core
Region of ␣-Synuclein Fibrils-Previous SDSL studies on ␣-synuclein have shown that same sites from different mole- cules come into close proximity (24). The accuracy of this analysis, however, was not sufficient to fully distinguish whether parallelism occurred between sheets (ϳ10 Å apart) or strands (4.7 Å apart), and it was not clear whether the spin-spin interactions arose from the proximity of two or more spin labels.
With sufficient spectral quality, such a distinction can be made by EPR as the close proximity between same sites in multiple parallel strands can give rise to characteristic exchange narrowed, single line EPR spectra (35,36). Because exchange narrowing is defined by single-line EPR spectra (37), the presence of hyperfine (outer) peaks can easily obscure this spectral line shape. To optimize spectral quality and minimize components from unpolymerized protein or other background labeling (possibly due to codon mistranslation (38)), we employed a C-terminal truncation mutant of ␣-synuclein, containing residues 1-115. The highly negatively charged C terminus of ␣-synuclein remained unstructured in the fibril and was inhibitory to fibril formation in vivo and in vitro (23-26, 39 -46). In agreement with these previous studies, we found that the C-terminal truncation mutant formed fibrils much more readily. According to thioflavin T fluorescence and far-UV CD spectroscopy, fibrils formed in ϳ8 h as compared with 48 h for the full-length protein (data not shown). Although the truncation mutant had a greater tendency to aggregate laterally, its fibrils had a highly similar morphology to that of wild-type ␣-synuclein, giving rise to fibril diameters of 6 -12 nm; this is in agreement with previously published results for wild-type ␣-synuclein as well as for C-terminal truncation mutants (40,41). We therefore generated singly R1-labeled 115ter deriva-tives and used EPR spectroscopy to investigate their structures in both soluble and fibrillar forms.
Spectra of all 115ter derivatives in solution were highly similar to each other. As illustrated with the representative examples of 25R1, 52R1, and 103R1 ( Fig. 2A, plotted at ϫ1/10 scale), the spectra have very sharp and narrowly spaced lines. These spectral features arise from a high degree of mobility (R1 rotational correlation time in the subnanosecond time scale), and are in agreement with the previously noted disordered structure in solution (24,(47)(48)(49).
As shown in Fig. 2, B-D, fibril formation induced significant spectral changes. Spectra derived from sites within the N terminus ( Fig. 2B: R1-labeled positions 25, 30, and 32) have multiple components of varying mobility, which is in agreement with the notion that the structure in this region is likely to be heterogeneous (24,25). While spectra from the C terminus ( Fig.  2D, R1-labeled positions 103, 108, and 109) are also heterogeneous, yet are dominated by sharp lines of high mobility, these components are still broader than those from the soluble protein (for example, compare position 103 spectra between 2A and 2D). Overall, the data from the C-terminal sites are consist-FIGURE 2. EPR spectra of R1-labeled ␣-synuclein (115ter) in soluble and fibrillar forms. A, EPR spectra of freshly prepared, predominantly monomeric spin-labeled derivatives harboring R1 at the indicated positions. B-D, EPR spectra of spin-labeled derivatives in fibrillar form. Spin labels were introduced at the indicated positions in the N-terminal region (B), in the core region (C), or at the C-terminal sites (D). All EPR spectra were obtained at room temperature using a 150 Gauss scan width and were normalized to the same number of spins. Because of the large difference in amplitude, the spectra for soluble ␣-synuclein (A) are shown at reduced size (ϫ1/10). FIGURE 3. Spin-dilution experiments of ␣-synuclein (115ter) fibrils containing varying mixtures of R1 and R1 labels at position 52. A, overlay of fibril first-derivative spectra from proteins labeled with the following percentages of R1: 100% (blue), 90% (cyan), 75% (magenta), 40% (red), 25% (green), and 15% (black). The remainder was labeled with R1Ј. All first-derivative EPR spectra were obtained at room temperature using a scan width of 150 Gauss and were normalized to the same number of spins. While the spectral lines are dominated by exchange narrowing at high R1 percentages, lower percentages indicate dipolar broadening. B, overlay of fibril absorption spectra generated from normalized first-derivative EPR spectra (from panel A) by integration. C, central line amplitudes of normalized EPR spectra as a function of R1 percentage (black dots connected by solid line). The dashed line schematically illustrates the continuous decrease in amplitude that would be observed, if only dipolar coupling were present.

␣-Synuclein Fibril Structure Analysis
ent with the previously reported presence of highly dynamic structure in this region (24 -26).
Most spectra derived from sites located within the core region ( Fig. 2C: R1-labeled positions 52, 60, 71, 80, and 90) are almost completely free of hyperfine lines (the remaining outer peaks amounts to less than 0.3% of the signal, as estimated by spectral simulations). The single-line EPR spectra obtained from these sites are reminiscent of those previously recorded from tau fibrils and clearly indicate the presence of spin exchange narrowing. Similar results were also obtained when R1-labeled 115ter derivatives were seeded with full-length wild-type ␣-synuclein fibrils (data not shown), suggesting that the full-length fibrils and 115ter ␣-synuclein adopt similar fibrillar structures. Thus, these data argue that, ␣-synuclein fibrils must generate a highly specific structure in which multiple, equivalent residues from different polypeptide chains come into direct contact.
To further investigate the effects of spin exchange narrowing, and to illustrate that such exchange narrowing requires simultaneous contact between multiple spin labels, we performed a series of spin-dilution experiments. In this set of experiments, fibrils were assembled from protein labeled at position 52 with various mixtures of R1 and its diamagnetic analogue, R1Ј (Fig. 1). The rationale for this spin-dilution experiment is that the presence of R1Ј will increase the distance between R1 groups and thereby decrease the effects of spinspin interactions, in particular that of exchange narrowing. As shown in Fig. 3A (the integrals/absorption spectra are shown in Fig. 3B), increasing dilution with R1Ј successively promotes the formation of three-line EPR spectra with clear hyperfine structure and the resulting spectral features that are indicative of strong immobilization. These data clearly show that the singleline EPR spectra observed for the fully R1-labeled fibrils are caused by spin-spin interactions.
Next, we systematically investigated the effect of spin dilution on the EPR spectra of 52R1/R1Ј fibrils (Fig. 3A) by recording the normalized spectral amplitude as a function of spin dilution (Fig. 3C, black dots connected by solid line). At lower percentages of R1 (Յ50%) the dipolar spin-spin interaction increases with increasing amounts of R1 and causes progressive line and spectral broadening as well as a loss in signal amplitude. This behavior is typically observed for dipolar broadening (50,51). The continued reduction in signal amplitude that one might expect for higher percentages of R1 in the case of purely dipolar interaction is schematically illustrated by the dashed FIGURE 4. Quantitative analysis of dipolar spin-spin interaction at low percentages of R1. Panel A shows the data and analysis from fibrils of ␣-synuclein (115ter) derivatives labeled at the indicated positions. EPR spectra of frozen samples (233K) labeled with 40% R1 (red traces) and 10% R1 (black traces; 15% in the case of 79R1) are compared in the left column. The red traces in the center column were obtained by deconvolution of the aforementioned spectra according to previously published methods (34). By fitting this experimentally obtained broadening function to a set of Pake broadening functions (black traces in center column), it was possible to convert the underlying dipolar interaction into the distance distributions given in the right column. The correspondence between the peaks and shoulders in the broadening functions and distance is illustrated by the arrows shown in the 90R1 derivative. The green spectra (left column) were obtained by applying the Pake broadening function (center column, black trace) to the black spectra (left column). The line shape for the green, Pake pattern-broadened spectrum coincides well with the experimentally observed spectrum for 40% R1 (red traces, left column), indicating that the set of Pake functions adequately describes the dipolar broadening. Panel B represents an analogous distance analysis using EPR spectra obtained previously (35) for spin-labeled tau fibrils labeled at position 308. Compared are data from 10 and 25% R1-labeled proteins. All EPR spectra are shown at a scan width of 200 Gauss and are normalized to the same number of spins. AUGUST 24, 2007 • VOLUME 282 • NUMBER 34 line in Fig. 3C. In the present case, however, the EPR spectra become narrower at higher percentages of R1, lose the hyperfine peaks and some of the dipolar broadening, and consequently regain in amplitude (Fig. 3C). Thus, exchange narrowing becomes more predominant as more and more spin labels are coming together.

␣-Synuclein Fibril Structure Analysis
It has been well established that dipolar broadening interactions can be described by a Pake pattern-type broadening function, which can also be used to obtain interspin distances (34,(52)(53)(54). We therefore tested whether such an analysis could also be used to describe the broadening that occurs at low percentages of R1 (Fig. 4). Using the method described by Altenbach et al. (34), we deconvoluted the spectra of ␣-synuclein fibrils that were obtained at 40% R1 with those obtained at lower R1 percentages (10 or 15%; Fig. 4A, left column) to directly obtain a broadening function that is model-independent (Fig. 4A, red trace in center column). The broadening functions for all of these derivatives are similar and indicate a broad range of distances. The defined peaks and shoulders suggest that the range of distances is made up of subsets of distances. Fitting to a weighted sum of Pake functions did indeed reveal that these peaks correspond to distances of Ͻ7 Å, ϳ10 Å, ϳ14 Å, and ϳ19 Å for all of the derivatives (Fig. 4A, right column). To further test the validity of the fit, the spectra obtained at low R1 percentages were convoluted with the fitted Pake functions. Although minor deviations could, in principle, arise from residual exchange narrowing or orientation effects (54), the resulting spectra (Fig. 4A, green traces, left column) overlay well with those obtained at 40% R1, which shows that the Pake functions FIGURE 5. EPR spectra of ␣-synuclein (115ter) fibrils containing single labels at positions 34 -101. Spectra from fibrils labeled with 100% R1 are shown in red. Spectra from samples labeled with a mixture of 25% R1 and 75% R1Ј are shown in green. Gray boxes highlight sites with significant hyperfine splitting in spectra from 100% R1-label. All EPR spectra were obtained at room temperature using a scan width of 150 Gauss, and were normalized to the same number of spins.

␣-Synuclein Fibril Structure Analysis
appropriately describe the dipolar broadening under the present conditions. Because we had previously observed exchange narrowing in tau fibrils, we next tested whether the same approach could be applied to spin dilutions for tau fibrils. As shown for tau fibrils labeled at position 308, analogous results were obtained (Fig. 4B). Interestingly, the distance peaks for the ␣-synuclein and tau fibrils correspond approximately to the distance between two, three, four, and five ␤-strands measured by fiber diffraction studies (23,55,56).
The data presented in Fig. 2C and Figs. 3 and 4 support the idea of a parallel arrangement in which same residues come into close contact. In order to investigate whether this behavior applied to all residues in the core, we performed a systematic nitroxide scanning experiment in which each residue between positions 34 and 101 was replaced by R1, one amino acid at a time. Fibril formation from these derivatives was verified by EM imaging and their structural properties were investigated by EPR spectroscopy. As shown in Fig. 5, strongly exchange-narrowed EPR spectra were observed for nearly all sites (Fig. 5, red spectra). These data show that, in ␣-synuclein fibrils, extended regions are arranged in a parallel, in-register manner that requires multiple molecules to stack on top of each other.
Despite the strong prevalence of exchange-narrowed spectra, a few sites exhibited a pronounced hyperfine structure (outer peaks), suggesting that they do not take up a stacked structure. These spectral features were most notable at the beginning or end of the scan (34, 99 -101), as well as in a region at the center of the scan (62-67) (Fig. 5, gray boxes). To a much lesser extent, a hyperfine structure is also present in the spectra of other sites (35, 36, 45, 47-49, 58 -59, 85, 94), but the overall line shape for these sites is still dominated by exchange narrowing (ϳ90% as estimated by spectral simulations).
In an effort to test whether exchange-narrowed EPR spectra could be observed for sites outside the core region, we generated nine additional derivatives harboring spin labels at various N-terminal sites. None of the fibrils from any of these derivatives gave evidence of significant exchange narrowing (Fig. 6, red spectra). Thus, unlike most regions of the fibril core, the Nor C-terminal regions do not have the same pronounced stacked, in-register organization.
Structural Features of ␣-Synuclein Fibrils from R1 Mobility and Accessibility-R1 mobility has been shown to be a sensitive indicator of local structure. Therefore, we sought to use R1 mobility to obtain additional structural information for ␣-synuclein fibrils. Because the EPR spectra from fully labeled fibrils are affected by mobility as well as by spin-spin interactions (30), we used the spin-dilution approach to diminish the effects of spin-spin interactions to focus on mobility. For each of the derivatives, we grew fibrils from a mixture of 25% R1 and 75% R1Ј, and recorded their EPR spectra (Figs. 5 and 6, green spectra). Under these conditions, mobility can be expressed by the semi-quantitative parameter, the inverse central line width (⌬H 0 Ϫ1 ) (57). As shown in Fig. 7A, there is a pronounced difference in ⌬H 0 Ϫ1 between the more mobile sites within the Nand C-terminal regions (Fig. 7A, gray boxes) and the strongly immobilized sites in the core region of the fibrils. Nearly all sites between residues 36 and 98 have very low values that are comparable to those observed for the fibril cores of tau and A␤ (35,58). Thus, as in those proteins, the fibril core of ␣-synuclein appears to be tightly packed. The most notable exception is a region encompassing residues 62-67 (Fig. 7A, gray box). Mobility values in this region are much higher, suggesting a less ordered and less tightly packed local structure. Residues 83-87 and residues 46, 58 also display elevated mobility, but to a much lesser effect. Sites near the beginning (34 -35) or the end (99 -101) of the core region seem to be located in transitional regions, as their mobility gradually transitions from low mobility in the core to higher mobility in the N and C termini (Fig. 7A,  gray boxes). Overall, these mobility values correlate well with the existence of exchange narrowing, because sites that exhibit exchange narrowing in the fully labeled state are generally of lowest mobility.
Although ␣-synuclein fibrils have significant ␤-sheet content, the mobility data do not exhibit regions of strong periodicity. For soluble globular proteins, a periodicity of two is commonly observed in ␤-sheets wherein exposed and buried residues alternate. Such a structure requires that one side of a sheet is solvent-exposed while the other is buried. The lack of periodicity in the ␣-synuclein fibril core therefore indicates the formation of sheets with packing interactions on both sides (possibly because of a combination of intrasheet, intersheet, and interfilament interactions). It is possible, however, that a facet of ␤-sheets in ␣-synuclein fibrils may be more surfaceexposed as well as more accessible to O 2 . To test this idea, we equilibrated the fibrils labeled with 25% R1 with pure O 2 (at atmospheric pressure, see "Experimental Procedures") and determined the oxygen accessibility (expressed as (O 2 )). As shown in Fig. 7A (red triangles, right Y-axis), the same regions that have elevated mobility (N and C termini, as well as residues 62-67) also have the greatest accessibilities. Our data illustrated that these more loosely packed regions are more readily accessible to O 2 . Although accessibility is generally low at most sites in the core region, we observed small periodic oscillations with a periodicity of two in several regions. These regions include residues 35-40 (EGVLYV), 51-54 (GVAT), 69 -82 FIGURE 6. EPR spectra of ␣-synuclein (115ter) fibrils containing single labels at selected sites in the N-and C-terminal regions. Spectra from fibrils labeled with 100% R1 are shown in red. Spectra from samples labeled with a mixture of 25% R1 and 75% R1Ј are shown in green. All EPR spectra were obtained at room temperature using a scan width of 150 Gauss and normalized to the same number of spins. AUGUST 24, 2007 • VOLUME 282 • NUMBER 34 (AVVTGVTAVAQKTV), 83-87 (EGAGS), and 95-98 (VKKD) (green boxes and green dashed box in Fig. 7A; residues facing outward are highlighted in aforementioned bold underlined letters), indicating the potential of a ␤-sheet structure in these regions (also see "Discussion").

DISCUSSION
The goal of the present study was to obtain more detailed structural information on the core region of ␣-synuclein fibrils. Toward this end, we generated 83 spin-labeled derivatives, including a complete nitroxide scan from positions 30 to 101, and studied the structural features of these derivatives using EPR spectroscopy.
Mobility and accessibility analyses revealed a tightly packed region extending from residue 36 to residue 98, with only few localized areas of elevated mobility and accessibility. Importantly, most sites within this range exhibited exchange-narrowed EPR spectra that were dominated by single lines with little or no detectable outer peaks. As illustrated in the spin-dilution experiments, these spectral lines were present only under conditions in which a large number of spin labels came into contact with each other. The close proximity of two labels at a time can give rise to an exchange interaction that causes a five-line EPR spectrum, and this has been observed for small biradicals and bilabeled peptides with nitroxides in close proximity (59,60). Single-line, exchange-narrowed EPR spectra, however, are very rare in spin-labeled proteins, as they require the orbital overlap of several labels. No single-line EPR spectra have been observed for proteins in which two R1 labels were in close proximity. This includes a pair of spin labels that were within vander Waals contact in a crystal of T4 lysozyme (61). In this case, physical contact was made by the flanking methyl groups, resulting in a nitroxide-nitroxide distance of 8.1 Å. Although the close proximity was reflected by strong dipolar broadening in the EPR spectrum, there was no evidence of any exchange interaction (neither exchange narrowing nor a five-line EPR spectrum). Furthermore, studies in which three (51) or four (50) spin labels came into close contact did not reveal any significant exchange narrowing. In crystals of spin labels, however, wherein an indefinite number of nitroxides are stacked on top of each other, such exchange narrowing is commonly observed (62). The distance between the spin labels in such crystals as well as in ␣-synuclein must be close enough to allow averaging of the hyperfine interaction. Thus, the frequency of the exchange interaction must be much greater than that of hyperfine interaction (Ͼ Ͼ10 8 s Ϫ1 ). The Pake pattern-based convolution/deconvolution method of Fig. 4 indicates that multiple spin labels are tightly packed together at a distance close to that of the interstrand distance derived from fiber diffraction (4.7-4.8 Å). The distance range of continuous wave EPR spectroscopy (Յ20Å) was sufficient to measure distances between spin labels located on up to five strands apart from each other. While larger distances could not be detected using the current methodology, the data are nevertheless consistent with a model Green shading indicates regions in which (O 2 ) exhibits a periodicity of 2 (at least four consecutive residues). The top panel gives the ␣-synuclein sequence from residues 35 to 100 using the aforementioned color code. B, schematic summary of EPR data. The top panel indicates the location of sites that have significant hyperfine structure (at more than four consecutive sites) in fibrils labeled with 100% R1. The panels below highlight the location of residues with elevated mobility or O 2 accessibility as described in A. The bottom panel highlights the location of regions in which the O 2 accessibility oscillates with a factor of 2. A periodicity of 2 indicates that protected and more exposed sites alternate. Such a periodicity would be expected in the case of ␤-sheets that are preferentially more exposed on one side. No obvious periodicity would be expected in sheets in which both sides are protected. Thus, the lack of periodicity in parts of the core region does not necessarily indicate the absence of ␤-structure.
wherein an indefinite number of spin labels are stacked along the fibril axis (Fig. 8).
In summary, the EPR spectra demonstrate that ␣-synuclein fibrils must be arranged in a parallel, in-register structure in which multiple polypeptides stack on top of each other. Overall, these data are in good agreement with the emerging view that parallel arrangement is a rather common feature shared by a number of fibrils; these include A␤, IAPP, a short fragment of ␤2-microglobulin, and the yeast prions Sup35p and Ure2p 10 -39, as well as crystals of short peptide fragments from amyloidogenic proteins (31,58,(63)(64)(65)(66)(67)(68). Given the fact that ␤-strands within the core region of ␣-synuclein (and other amyloid) fibrils are arranged perpendicular to the fibril axis (Fig. 8), a parallel, in-register structure requires that each layer of the fibril (every 4.7-4.8 Å, Fig. 8) must contain a new polypeptide molecule. Such a structure maximizes intermolecular contact surface, a feature that is likely to aid in the highly specific fibril propagation (24,35).
Another unique feature of fibrils with parallel, in-register structure is that their stability is strongly affected by the packing interactions between same residues that stack on top of each other (36). A recent study on tau fibrils demonstrated that the stacking interactions of charged residues are strongly destabilizing, while the stacking of ␤-branched residues, such as Ile and Val, is highly stabilizing. In this respect, it is noteworthy that ␣-synuclein contains a repeat region found to be critically important for fibril formation (residues 71-82) (69,70). This region, which is not present in the less amyloidogenic ␤-synuclein, differs from other repeats in that it is devoid of charged residues and instead contains a high proportion of hydrophobic ␤-branched residues (see sequence in Fig. 7A). Thus, the increased amyloidogenic nature conferred by this region can readily be rationalized by the extensive interactions that occur between same residues in parallel, in-register structures.
Although the core region of ␣-synuclein fibrils is characterized by pronounced stacking interactions and strong immobilization, there are some regions that do not experience such a structure. The most pronounced lack of stacking interactions and highest mobility were observed for the region around residues 62-67 (Fig. 5, gray box; Fig. 7A, gray box; Fig. 7B, top two panels). These data suggest that this region, which coincides with the beginning of the NAC region, represents a less tightly packed region that is probably in a loop or turn conformation. Increased mobility and loss of stacking interactions were also observed at residues 34, 35, and 99, which are located at the Nand C-terminal boundaries of the core region (Fig. 7A, gray  boxes; Fig. 7B, top two panels).
While mobility information clearly revealed that the aforementioned regions are of higher mobility, it did not directly reveal the location of the ␤-sheet structure. Such an assignment can readily be accomplished when a given ␤-sheet is exposed on one side and buried on the other side (28). The fact that no such periodicity could be observed in ␣-synuclein fibrils indicates that significant packing interactions must be present on all sides of the ␤-sheets within the fibril. By performing accessibility measurements using 100% molecular O 2 , however, we were able to increase sensitivity. The accessibility data served to identify several regions with a periodicity of two, which is characteristic of ␤-structure (Fig. 7A, green boxes: 35-40, 51-54, 69 -87, and 95-98; Fig. 7B, bottom panel). The most extensive periodicity was found for residues 69 -87, which contain the key repeat region that is absent in ␤-synuclein. ssNMR has also suggested ␤-structure in this region, although there was indication that its C-terminal end might be in a bend or turn conformation (26). It should be noted that we observed some, albeit a rather small, increase in overall mobility for residues 83-87 (Fig. 7A, dashed green box), as well as the presence of some hyperfine structure at residue 85. These observations would be consistent with the formation of a more loosely stacked structure in this region. Although additional data are required to pinpoint the precise end of the ␤-structure, we can clearly define which face of the ␤-structure in this region is more exposed to O 2 ( 69 AVVTGVTAVAQKTVEGAGS 87 , highlighted in the aforementioned bold underlined letters).
It has recently been suggested that the N-terminal portion of the core region could have different structures depending upon fibril morphology (26). While structural heterogeneity might well exist in this region, it is important to point out that exchange narrowing was observed throughout the core region. Thus, the parallel, in-register orientation must be maintained in all structures.
Collectively, our EPR data, together with information from other structural techniques, have provided significant structural constraints for ␣-synuclein fibrils. With additional interresidue distance information from ssNMR and continuous wave as well as pulsed EPR, it should ultimately be possible to obtain reliable three-dimensional models or structures. FIGURE 8. Schematic illustration of ␣-synuclein fibril structure highlighting the parallel, in-register arrangement of multiple strands. The outer cylinder schematically illustrates the overall fibril, and the inner cylinders represent the individual filaments within the fibril. Strands within one of those filaments are represented by green arrows. According to fiber diffraction (23), ␣-synuclein takes up a cross-␤ structure in which ␤-strands run perpendicular to the fiber axis (4.7 ϳ 4.8 Å apart). The EPR data show that a spin label introduced at a given site within the core region (schematically indicated by orange circles) stacks on top of equivalent sites in neighboring polypeptides. The close contact between same residues at sites throughout the core region demonstrates a parallel, in-register structure in which multiple strands run parallel to each other. Importantly, each layer contains a new molecule. Such a structure maximizes intermolecular contact surface, a feature that is likely to aid in the highly specific fibril propagation.